That was interesting. And the quadratic benchmark listed after that highlights how hyperscan wins in an unfair way - it just implements different semantics, so

That was interesting. And the quadratic benchmark listed after that highlights how hyperscan wins in an unfair way - it just implements different semantics, so it's not comparing the same feature. (I see now, after posting, you've mentioned it in another comment.)